Striatal Neurons Expressing D1 and D2 Receptors are Morphologically Distinct and Differently Affected by Dopamine Denervation in Mice

The loss of nigrostriatal dopamine neurons in Parkinson’s disease induces a reduction in the number of dendritic spines on medium spiny neurons (MSNs) of the striatum expressing D1 or D2 dopamine receptor. Consequences on MSNs expressing both receptors (D1/D2 MSNs) are currently unknown. We looked for changes induced by dopamine denervation in the density, regional distribution and morphological features of D1/D2 MSNs, by comparing 6-OHDA-lesioned double BAC transgenic mice (Drd1a-tdTomato/Drd2-EGFP) to sham-lesioned animals. D1/D2 MSNs are uniformly distributed throughout the dorsal striatum (1.9% of MSNs). In contrast, they are heterogeneously distributed and more numerous in the ventral striatum (14.6% in the shell and 7.3% in the core). Compared to D1 and D2 MSNs, D1/D2 MSNs are endowed with a smaller cell body and a less profusely arborized dendritic tree with less dendritic spines. The dendritic spine density of D1/D2 MSNs, but also of D1 and D2 MSNs, is significantly reduced in 6-OHDA-lesioned mice. In contrast to D1 and D2 MSNs, the extent of dendritic arborization of D1/D2 MSNs appears unaltered in 6-OHDA-lesioned mice. Our data indicate that D1/D2 MSNs in the mouse striatum form a distinct neuronal population that is affected differently by dopamine deafferentation that characterizes Parkinson’s disease.

The striatum is the main input structure as well as the largest integrative component of the basal ganglia. It receives a multitude of neurochemical inputs that are largely processed by striatal projection neurons. At the somatodendritic level, these cells form a rather morphologically homogeneous population, each element being endowed with a medium-sized cell body and typical spiny dendrites. These so-called medium spiny neurons (MSNs) all use γ -aminobutyric acid (GABA) as a neurotransmitter and represent approximately 90-95% of the striatal neuronal population in rodents 1 . Despite their morphological similarities, MSNs can be divided into two subpopulations based on their neurochemical content and axonal projection sites. Roughly half of the MSNs express dopamine (DA) receptor of the D 1 type and contain the neuropeptides substance P (SP) and dynorphin (DYN). They innervate mainly the substantia nigra pars reticulata and the entopeduncular nucleus (rodent homologue of primate internal pallidum) and form the so-called "direct pathway". The other half of the MSNs expresses DA receptor of the D 2 type and contains the neuropeptide enkephalin (ENK). Their axon arborizes principally in the pallidum (rodent homologue of primate external pallidum) and forms the first segment of the so-called "indirect pathway" [2][3][4][5] . However, it is worth noting that single-axon tracing experiments in rodents 6 and primates 7 indicate that most striatofugal axons arborize into the three main striatal targets. To this regards, it has recently been shown that both D 1 and D 2 MSNs located in the nucleus accumbens (Acb) can either inhibit or disinhibit thalamic activity depending on their projection pattern and not on their genetic characteristics 8 .
The D 1 and D 2 receptors are reportedly co-expressed in a certain proportion of MSNs, but the size of such D 1 / D 2 subpopulation is still a matter of controversy. Earlier studies undertaken with in situ hybridization methods, immunohistochemistry or reverse transcription polymerase chain reaction have reported high percentages of striatal neurons expressing both DA receptors 3,[9][10][11][12][13][14][15][16][17] , but a much smaller number of D 1 /D 2 MSNs was detected in transgenic mice expressing fluorescent reporters for D 1 or D 2 When co-expressed by MSNs, the D 1 and the D 2 receptors are reportedly able to form heteromers, the activation of which can lead to a distinct intracellular signalling pathway [22][23][24][25][26] . The independent activation of the D 1 or the D 2 DA receptor is known to differentially regulate cyclic-AMP activity, respectively leading to the activation (D 1 coupled to G s ) or inhibition (D 2 coupled to G i ) of MSNs 27,28 . In contrast, the activation of D 1 /D 2 heteromers would result in a distinct phospholipase C-mediated calcium signalling through activation of G q protein [22][23][24][25] . However, while co-expression of D 1 and D 2 receptors is well accepted, the existence of D 1 /D 2 heteromers in vivo remains controversial 29,30 .
The fate of MSNs expressing D 1 or D 2 DA receptor in the context of DA striatal deafferentation that characterizes Parkinson's disease (PD) has already been studied. In a mouse model of PD, lesion of the striatal DA input was shown to induce spine pruning, principally on the D 2 MSNs 31,32 and less markedly on the D 1 MSNs [31][32][33][34][35] . Such spine pruning appears to be a highly plastic phenomenon since the reduction of spine density observed on the D 2 MSNs, but not that on D 1 MSNs, could be restored by long-term administration of L-Dopa 33,35 . Reduction in the number of dendritic spine of striatal MSNs has also been reported in non-human primate model of PD 36 and in PD patients [37][38][39] . Surprisingly, the fate of MSNs expressing both the D 1 and the D 2 DA receptors (the D 1 / D 2 MSNs) has never been investigated in PD condition. Therefore, we have designed a study to provide the first detailed description of changes induced by DA denervation in the density, regional distribution and fine morphological characteristics of dendritic processes of D 1 /D 2 MSNs that populate the dorsal striatum and the Acb, the main component of the ventral striatum, of mice. Using stereological approaches and single-neuronal injections performed on striatal sections from sham and 6-hydroxydopamine (6-OHDA)-lesioned double BAC transgenic mice (Drd1a-tdTomato/Drd2-EGFP), we show that the D 1 /D 2 MSNs are affected differently than the D 1 and D 2 MSNs by striatal DA deafferentation that characterizes PD.

Results
Unilateral 6-OHDA injections cause severe TH+ cell loss in the SNc and VTA, significant DA depletion in the striatum and in the Acb and spontaneous ipsilateral rotations. Immunolabeling of the striatum and the substantia nigra pars compacta (SNc) for the DA transporter (DAT) and tyrosine hydroxylase (TH) indicate a severe DA lesion caused by 6-OHDA unilateral injections performed in the medial forebrain bundle (Fig. 1). Counts of TH + cell bodies in the midbrain show a more severe DA cell loss in the SNc compared to the VTA (80.7 ± 7.5% decrease in SNc vs. 48.9 ± 9.2% decrease in VTA, compared to control side, Fig. 1a,b).
LI-COR ® slide scanner measurements indicate an average of 81.3% decrease of TH and 87.5% of DAT immunoreactivity in the dorsal striatum when compared to control side as well as a 80.1% reduction of TH and a 79.3% decrease of DAT immunoreactivity in the Acb, when compared to control side (Fig. 1c,d). Behavioral assessments show a significant preference for spontaneous rotations ipsilateral to the lesioned side in mice that were unilaterally injected with 6-OHDA (57.1 ± 6.4 spontaneous ipsilateral rotations/10 min vs. 1.1 ± 0. 8 contralateral rotations, P < 0.0001, Fig. 1e).
The D 1 /D 2 MSNs contain dynorphin but not enkephalin. Examination of ENK and DYNimmunostained sections from colchicine-treated mice reveals that, as expected, D 1 MSNs are immunoreactive for DYN but not for ENK. In contrast, D 2 MSNs contain ENK but are devoid of DYN. In regard to D 1 /D 2 MSNs, they are immunopositive for DYN but immunonegative for ENK (Fig. 2). D 1 /D 2 MSNs are homogeneously distributed throughout the dorsal striatum but heterogeneously scattered in the nucleus accumbens. The regional distribution of D 1 , D 2 and D 1 /D 2 MSNs in the dorsal striatum of sham mice was estimated stereologically. The overall densities of D 1 and D 2 MSNs are 108 ± 5 and 95 ± 4 × 10 3 cells/mm 3 , representing respectively 52.2 ± 1.1% and 45.9 ± 1.1% of the total MSNs population of the striatum. In contrast, the D 1 /D 2 MSNs have a much lower density with only 3.8 ± 0.3 × 10 3 cells/mm 3 (P < 0.0001), representing 1.9 ± 0.2% of the striatal MSNs. Although not statistically significant (P = 0.2341), the density of the D 1 /D 2 MSNs in the ventromedial sector of the striatum was lower in the post-commissural striatum (1.8 ± 0.4 × 10 3 cells/mm 3 ) than in the pre-commissural striatum (4.3 ± 0.3 × 10 3 cells/mm 3 ). Statistical evaluations (ANOVA) of neuronal densities in different striatal regions reveal no statistical differences (Fig. 3), supporting the homogeneous regional distribution of the D 1 , D 2 and D 1 /D 2 MSNs throughout the dorsal striatum. Occasionally, some D 1 /D 2 MSNs can be seen to form small clusters of 2-3 cells at different striatal levels. Assessment of the distribution of D 1 /D 2 MSNs in striosomes and matrix striatal compartments, as delineated on immunostained sections for μ -opioid receptor, reveals no significant difference, neither at the pre-commissural level (P = 0.4857), nor at the post-commissural level (P = 0.8286, Fig. S1). Our quantitative analyses reveal that the D 1 /D 2 MSNs are more heterogeneously distributed in the Acb than in the dorsal striatum. As shown in Fig. 4d, a dense region located in the medial part of the shell of the Acb was almost entirely composed of the D 1 /D 2 type of MSNs. A closer examination of the lateral striatal stripe indicates no significant difference regarding the density of the D 1 /D 2 MSNs compared to other regions of the lateral area of the shell of the Acb. The density of the D 1 /D 2 MSNs is overall significantly higher in the Acb compared to the striatum. This difference becomes statistically significant in the shell compartment of the Acb with 45.0 ± 10.9 × 10 3 cells/mm 3 compared to 3.8 ± 0.3 × 10 3 cells/mm 3 in the dorsal striatum (P = 0.0098, Fig. 3c). Our stereological estimations indicate that the D 1 /D 2 MSNs represent 14.6 ± 3.0% of the MSN population in the shell and 7.3 ± 2.2% in the core compartment of the Acb, percentages that are significantly higher than what was noted in the dorsal striatum (1.9 ± 0.2%, P = 0.0098). Interestingly, the density of the D 1 MSNs is also higher in the shell of the Acb when compared to the dorsal striatum (180.0 ± 13.6 vs. 108.2 ± 5.4 × 10 3 cells/mm 3 , P = 0.0244, Fig. 3a) whereas no difference is noted regarding the density of the D 2 MSNs between the dorsal striatum and the Acb. A higher density of the D 1 and the D 1 /D 2 MSNs in the Acb is congruent with an overall higher density of all MSNs in the shell (298.6 ± 15.0) and the core (267.9 ± 24.5 cells/ mm 3 ) compartments of the Acb compared to the dorsal striatum (123.5 ± 5.4 × 10 3 cells/mm 3 ).  MSNs (P = 0.0009). The reconstruction of somatodendritic domains of Lucifer yellow-filled MSNs reveals that the total dendritic length of D 1 /D 2 MSNs is also smaller than that of the D 1 and the D 2 MSNs. The mean total dendritic length for the D 1 /D 2 MSNs is 0.75 ± 0.06 mm compared to 1.48 ± 0.13 mm for D 1 (P < 0.0001) and 1.08 ± 0.08 mm for D 2 MSNs (P = 0.0296, Fig. 5a). We also noted that the dendritic arborization of the D 1 MSNs is significantly longer than that of the D 2 MSNs (P = 0.0089). The dendritic arborization of the D 1 /D 2 MSNs is significantly less profuse than that of the D 1 and D 2 MSNs, as indicated by a smaller number of dendritic branching points (9.5 ± 0.8 branching points) compared to the D 1 (16.6 ± 1.5 branching points, P = 0.0015) and the D 2 MSNs (15.0 ± 1.4 branching points, P = 0.0087, Fig. 5b). D 1 /D 2 MSN dendrites harbor fewer spines than the two other types of MSNs. By dividing the number of spines by the total dendritic length for each reconstructed neuron in sham animals, we were able to evaluate the overall spine density for each of the three types of striatal MSNs. Our data reveal that the D 1 /D 2 MSNs have a 37% lower spine density (4.0 ± 0.2 spines/10 μ m) than the D 1 (6.4 ± 0.3 spines/10 μ m, P < 0.0001) and the D 2 (6.6 ± 0.2 spines/10 μ m, P < 0.0001) MSNs (Fig. 5c). A Sholl analysis performed on all reconstructed neurons indicates that this lower spine density is maintained throughout the entire dendritic extent of the D 1 / D 2 MSNs, the difference being statistically significant on a distance ranging between 45 and 105 μ m from their parent cell bodies (Fig. 5d).
The extent of D 1 /D 2 MSN dendritic arborization is unaffected by 6-OHDA lesion, in contrast to that of D 1 and D 2 MSNs. Statistical comparison between sham and 6-OHDA-lesioned mice in regard to the extent of somatodendritic domain belonging to the D 1 , D 2 and D 1 /D 2 MSNs in the dorsal striatum indicates that the total dendritic length of the D 1 MSNs was reduced in the DA-depleted striatum by 60% (0.60 ± 0.04 vs. 1.48 ± 0.13 mm, P < 0.0001, Fig. 6a) and by 28% for the D 2 MSNs (0.78 ± 0.09 vs. 1.08 ± 0.08 mm, P = 0.0191, Fig. 6b). Surprisingly, no significant differences between sham and 6-OHDA-lesioned mice were observed regarding the total dendritic length of the D 1 /D 2 MSNs (0.61 ± 0.05 vs. 0.75 ± 0.07 mm, P = 0.6356, Fig. 6c). Accordingly, in 6-OHDA-lesioned mice, a lower number of dendritic branching points were observed for the D 1 MSNs (8.7 ± 0.7 vs. 16.6 ± 1.5 branching points, P < 0.0001) and the D 2 MSNs (10.7 ± 0.9 vs. 15.0 ± 1.4 branching points, P = 0.0120) but not for the D 1 /D 2 MSNs with 9.5 ± 0.8 branching points in both experimental groups (P > 0.9999). The spine density on D 1 /D 2 MSN dendrites is reduced in 6-OHDA-lesioned mice. Lesion of the striatal DA afferent projections leads to a significant decrease of spine density on dendrites belonging to the three types of MSNs. The overall spine density was 4.0 ± 0.4 spines/10 μ m in 6-OHDA compared to 6.4 ± 0.3 in sham (P < 0.0001) for the D 1 MSNs; 5.1 ± 0.3 spines/10 μ m compared to 6.6 ± 0.2 (P = 0.0018) for the D 2 MSNs and 3.0 ± 0.1 spines/10 μ m compared to 4.0 ± 0.2 (P = 0.0427) for the D 1 /D 2 MSNs. These reductions accounted for a 37.5% loss on the D 1 , 22.7% on the D 2 and 25.0% on the D 1 /D 2 MSNs. The Sholl analysis indicates that such lower spine density is maintained throughout the entire dendritic arborization of the three types of MSNs (Fig. 7).

Discussion
The present study provides the first detailed description of the morphological characteristics, density and regional distribution of D 1 /D 2 MSNs of the dorsal striatum and Acb in normal mice, as well as the first characterization of changes induced in this striatal subpopulation by striatal DA denervation. Our data gathered in normal animals reveal that the D 1 /D 2 MSNs are morphologically distinct from the D 1 and D 2 MSNs: they have a smaller cell body, a less profusely arborized dendritic tree with branches that bear fewer spines than those of the D 1 and D 2 MSNs. They are uniformly scattered throughout the striatum, where they represent approximately 2% of the total number of MSNs, but heterogeneously distributed and more abundant in the Acb, where their proportion ranged from 7 to 15% of all MSNs. In 6-OHDA-lesioned mice, the density and regional distribution of all 3 types of MSNs is essentially unaltered. In contrast to the D 1 and the D 2 MSNs, the D 1 /D 2 neurons do not show any significant reduction in the length of their dendritic arborization after intoxication with 6-OHDA. However, a reduction in dendritic spine density was noted in all three types of MSNs following DA depletion, but this pruning phenomenon was more prominent in the D 1 than in the D 2 or D 1 /D 2 MSNs. The significance of these results will now be discussed in the light of relevant literature.
It is now well established that co-expression of D 1 and D 2 receptors occurs in some striatal MSNs, but there is still some controversy regarding the relative importance of such a unique neuronal population. In the literature, the percentage of striatal MSNs that coexpress D 1 and D 2 ranges from low 2,10,11,24,40 , moderate 12,14,41 to nearly 100% 9 . Such substantial differences may be explained, on one hand, by the various methods and species used and, on the other hand, by the specificity of the antibodies or the in situ hybridization probes employed to detect D 1 and D 2 receptors. In the present study, depending on striatal sectors that were examined, we estimate that the D 1 /D 2 MSNs account for 0.8-2.4% of the total number of striatal MSNs, a proportion that agrees with figures reported in other studies conducted in BAC transgenic mice 18,20,42 .
Despite that they represent only 2% of the total MSNs population of the adult mice dorsal striatum, the D 1 /D 2 MSNs might play an important role in striatal functioning, as it is the case of striatal interneurons, which account only for 2-3% of striatal neurons in rodents 43,44 . Their presence throughout the dorsal striatum suggest that the D 1 /D 2 MSNs are involved in the sensorimotor and associative functions of the striatum, which are integrated mainly within the caudolateral and the rostromedial sector of the structure, respectively 45 . However, the prevalence of the D 1 /D 2 MSNs in the Acb indicates that these neurons are even more actively implied in the limbic aspect of striatal functioning. Indeed, we found the density of D 1 /D 2 MSNs in the Acb to be significantly higher than in the dorsal striatum, a finding that is congruent with data from previous studies conducted in transgenic mice 20,21,24,46,47 and by a higher number of D 1 /D 2 heteromer in the rat and monkey Acb 25,26 . The shell compartment of the Acb was significantly more enriched in D 1 /D 2 MSNs than the core compartment, supporting the notion that the latter is more similar to the dorsal striatum than the former, which has been described as a transition zone between the striatum and the extended amydgdala 48 . The fact that the D 1 /D 2 MSNs form the vast majority of MSNs in the medial area of the shell compartment, as noted here, is interesting since this medial region of the Acb shell is known to be involved in feeding behaviors [49][50][51][52] as well as in the response to noxious stimuli 52 . Whether or not the D 1 /D 2 MSNs present in the Acb play a role in the antidepressant and anxiolytic effects observed after disruption of the D 1 -D 2 complex 53 remains to be investigated.
Besides their difference in the expression of DA receptors, the two major types of striatal MSNs express different neuropeptides in both rodents and primates 2,14,54 , the D 1 MSNs containing SP and DYN while the D 2 MSNs are enriched with ENK 20 . As for the D 1 /D 2 MSNs, a previous report has suggested that they express both ENK and DYN in the rat 24 , whereas the present findings clearly show that they display immunoreactivity only for DYN. Such a discrepancy might reflect a species difference, but most likely results from a variation in the methodological approach. We used highly specific transgenic fluorescent reporters to identify the D 1 , D 2 and D 1 /D 2 striatal MSNs, whereas Perreault, et al. 24 employed antibodies against the D 1 and D 2 receptors, which are known to be also highly expressed in the striatal neuropil, a situation that might have hampered the proper identification of the various peptide expressing MSNs. Our data indicate that, in regard to neuropeptide content, the striatal D 1 / D 2 neurons in mice have more in common with the D 1 than with the D 2 MSNs. This first detailed report on the morphological organization of D 1 /D 2 MSNs reveals that these neurons have a smaller cell body and a shorter dendritic tree than their D 1 or D 2 counterparts. Our data also underline the less profuse dendritic aborization of the D 2 compared to the D 1 MSNs, a morphological feature that might explain, at least in part, why the D 2 MSNs are more excitable than D 1 MSNs 55 . Based on the morphology of their somatodendritic domains, it is tempting to speculate that the excitability and input resistance of the D 1 /D 2 MSNs would be higher than the D 1 or even the D 2 MSNs, but such a view needs to be confirmed by detailed investigations of electrophysiological properties of the D 1 /D 2 MSNs.
In addition to a smaller dendritic tree, the D 1 /D 2 MSNs also contain less dendritic spines than D 1 and D 2 MSNs. The head of dendritic spines is the preferential synaptic contact site of glutamatergic corticostriatal projections arising from the cerebral cortex and the intralaminar thalamic nuclei 56 . With their less profuse dendritic arborization, the D 1 /D 2 MSNs might receive significantly less glutamatergic input and thus be less vulnerable to excitotoxicity involved in different neuropathological conditions such as Huntington's disease. More importantly in the context of the present study is the spatial distribution of DA terminals in contact with different parts of the MSNs. Those terminals that contact the cell body and proximal dendritic shafts might produce a relatively non-specific effect mediated by the volumic release of DA 57 . In contrast, as suggested Freund and colleagues 58 , the major DA input that occurs on the necks of dendritic spines is likely to be much more selective since it could prevent the excitatory glutamatergic input to the same spines from reaching the dendritic shaft. One of the main functions of striatal DA release might be to alter the pattern of firing of striatal output neurons by regulating their input 58 .  Lesions with 6-OHDA in rats were shown to increase the expression of D 2 and decrease that of D 1 by MSNs 59 . Here we report that the density and regional distribution of D 1 /D 2 MSNs in mice striatum are unaltered by 6-OHDA lesions, suggesting that DA denervation does not alter the D 1 expression by D 2 MSNs or the D 2 expression by D 1 MSNs. However, it should be noted that the BAC transgenic reporter system used here does not fully allow to rule out the possibility that DA lesion may induce more subtle variations of DA receptor gene expression or receptor trafficking and degradation that could have remained undetected. The D 1 /D 2 MSNs do not display any reduction of their dendritic tree following 6-OHDA lesions, in contrast to D 1 and D 2 MSNs, but the D 1 /D 2 MSNs show a lower dendritic spine density in the DA-denervated striatum, as it is also the case for D 1 and D 2 MSNs. Reduction of the dendritic length of D 1 and D 2 MSNs following DA lesion is supported by observations gathered in mice 34,35 , monkeys 36 and PD patients [37][38][39] . However, some studies in rats 60 and mice 33 failed to demonstrate such a phenomenon. The time between DA lesion and animal sacrifice might explain such a discrepancy. The interval between 6-OHDA injection and animal perfusion in the latter two studies ranged from 4 to 5 weeks, whereas the delay was much longer (8 weeks) in the present study. In face of such differences, it is tempting to speculate that the reduction in dendritic length occurs after the spine loss has occurred, that is in the late stages of the disease.
We documented spine loss for the three types of striatal MSNs, in accordance with data obtained for D 1 and the D 2 MSNs in mice 33,35,61 , rats 60,62,63 , monkeys 36 and PD patients 37,38 . Interestingly, other studies have suggested that such spine loss may be restricted to the D 2 MSNs in mice 31,32 . As mentioned above, DA axons are ideally that were filled with Lucifer yellow in sham and 6-OHDA-lesioned mice. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 for sham vs. 6-OHDA by a Student's T-test (histograms) or Two-way ANOVA followed by Bonferroni's multiple comparison test (Sholl analysis). positioned on the dendritic spine neck to influence the effect of corticostriatal and thalamostriatal glutamatergic input 56 . It has been hypothesized that the loss of DA afferents may destabilize the morphological integrity of the spine, potentially leading to spine pruning, a phenomenon that might be the result of maladaptive calcium influx through the L-type calcium channels located on MSNs 31 . Moreover, evidence has recently been gathered regarding the implication of acetylcholine, the level of which is known to be increased in PD 64,65 , in spine pruning of the D 2 MSNs through the modulation of Kir2 channels, leading to an increase of dendritic excitability driven by the activation of M1 muscarinic receptor 32 .
The exact function of the D 1 /D 2 MSNs of the striatum remains elusive. Colocalization of the D 1 and D 2 DA receptors has been reported in axons located in various basal ganglia components 24 . Whether this observation indicates a distinct striatofugal pathway remains unclear 66 . However, the existence of such unique striatofugal projections would imply a complementary functional role of the D 1 /D 2 MSNs, working in concert with the D 1 and the D 2 MSNs for harmonious basal ganglia functioning. Single-axon tracing studies in rodents 6,67 and monkeys 7,68 have emphasized the highly collateralized nature of the striatofugal projections, challenging the concept of a simple dual (direct/indirect) striatofugal system. Whether the D 1 /D 2 MSNs contribute to these highly collateralized striatofugal projections is currently unknown and single-axon tracing of D 1 /D 2 MSNs is obviously needed to better appreciate their functional role in the basal ganglia functioning.

Animals.
This study was carried out on 25 double BAC transgenic mice (Drd1a-tdTomato/Drd2-EGFP) of 2 month old weighing between 20-30 g. Equal numbers of male and female were used. These D 1 /D 2 mice were generated by breeding B6SJLF1-D 1 tdTomato BAC transgenic mice 19 with C57Bl6J-D 2 -EGFP BAC animals 69 . They allow direct identification of the D 1 , D 2 and D 1 /D 2 MSNs (Fig. 8). In order to minimize over expression artifacts, all mice where heterozygous for each individual BAC transgene. Animals were housed under a 12 h light-dark cycle with water and food ad libitum. All procedures were approved by the Comité de Protection des Animaux de l'Université Laval, in accordance with the Canadian Council on Animal Care's Guide to the Care and Use of Experimental Animals (Ed2) and with the ARRIVE guidelines. Maximum efforts were made to minimize the number of animals used.

Stereotaxic injections. 6-OHDA unilateral injection and behavioral assessment. Nineteen
Drd1a-tdTomato/Drd2-EGFP transgenic mice received an intracerebral injection of 6-OHDA (catalog no. H4381; Sigma-Aldrich, Saint-Louis, MO, USA) in the right medial forebrain bundle (mfb). Approximately 30 minutes before 6-OHDA administration, mice received intraperitoneal injection of desipramine (25 mg/kg) diluted in saline (0.9%) at a concentration of 2 mg/mL. Mice were then anaesthetized using 2% isoflurane and their head were fixed in a stereotaxic apparatus. A hole was drilled and the following stereotaxic coordinates were aimed: anteroposterior (bregma) = − 1.2 mm; mediolateral = 1.1 mm; dorsoventral = − 5.0 mm, corresponding to the mfb, according to the mouse brain atlas of Franklin and Paxinos 70 . A glass micropipette of 35 μ m diameter at the tip containing a freshly prepared 6-OHDA solution was introduced in the mfb. The 6-OHDA was then pressure-injected and the micropipette was left in place for 2 min both prior and following the injection. The 6-OHDA was diluted in ascorbic acid (catalog no. A5960; Sigma-Aldrich) at a concentration of 6 μ g/μ L. A total volume of 0.25 μ L of 6-OHDA, corresponding to a dose of 1.5 μ g of the neurotoxin, was injected into the right mfb. The sham-lesioned group was composed of 4 mice that only received injections of the vehicle (0.02% ascorbic acid). After surgery, the skin was sutured and mice were allowed to recover.
Thirty days after surgery, mice from the two experimental groups were introduced in a large glass cylinder and spontaneous motor behavior was recorded during 10 min. Spontaneous rotations were counted by a blinded experimenter. Complete ipsilateral and contralateral rotations to the 6-OHDA-lesioned side were counted and used as behavioral indication of the severity of the DA lesion.
Fifty-six days after the 6-OHDA lesion, animals were deeply anesthetized with a mixture of ketamine (100 mg/kg) and xylazine (10 mg/kg). They were transcardially perfused with an initial wash of 40 mL of ice-cold sodium phosphate-buffered saline (PBS, 0.1 M; pH 7.4), followed by 150 mL of paraformaldehyde (PFA, 4% diluted in phosphate buffer). Brains were dissected out, post-fixed for 24 h in a 4% PFA solution and cut with a vibratome (model VT1200; Leica, Germany) into 50 μ m-thick coronal sections, which were serially collected in sodium phosphate-buffered saline (PBS, 0.1 M, pH 7.4). The pre and post-commissural parts of the striatum were cut at 50 μ m and used for immunohistochemistry and stereology whereas the commissural part was cut at 250 μ m for intracellular injections.
Bilateral colchicine injections. Two other double BAC transgenic mice received bilateral injections of colchicine (catalog no. C9754; Sigma-Aldrich) in the striatum, a drug that block axonal transport, allowing optimal immunostaining of neuropeptides contained in MSNs cell bodies. For intra-cerebral injections of colchicine, the following stereotaxic coordinates were used: anteroposterior (bregma) = 0.14 mm; mediolateral = 2.00 mm; dorsoventral = 3.20 mm, corresponding to the dorsal part of the striatum at the commissural level, according to the mouse brain atlas of Franklin and Paxinos 70 . One μ L of colchicine diluted at 23 mg/mL in saline was pressure-injected in each side of the brain. One week after injections, colchicine-injected mice were transcardially perfused, as described above. Their whole brains were dissected out and cut with a vibratome into 50 μ m-thick transverse sections.

Immunohistochemistry. TH and DAT immunohistochemistry.
To assess the extent of the DA lesion induced by 6-OHDA injections, one 50 μ m-thick section was selected from the pre-commissural striatum (0.14 mm from bregma) and from the SNc (− 3.52 mm from bregma), in each mouse. These sections were immunostained for TH, the catalytic enzyme of DA synthesis, by using a polyclonal antibody (catalog no. AB152; Scientific RepoRts | 7:41432 | DOI: 10.1038/srep41432 Millipore Corporation, Billerica, USA) raised in rabbit. Briefly, the free-floating sections were sequentially incubated in (i) a blocking solution of PBS containing 2% normal goat serum and 0.01% Triton X-100 (1 h, RT); (ii) the same solution containing a 1/1000 dilution of rabbit polyclonal antibody against TH (overnight, 4 °C); and (iii) a 1/500 solution of biotinylated goat anti-rabbit (catalog no. BA-1000; Vector Laboratories, Burlingame, CA, USA) diluted in the same blocking solution (2 h at room temperature (RT)). After rinses in PBS, sections were incubated for 1 h at RT in an avidin-biotin-peroxydase complex solution (Vector Laboratories) diluted 1/100 in the blocking solution. Sections were then rinsed and the bound peroxidase revealed by incubating the sections for 3 min at RT in a 0.025% solution of 3,3′ diaminobenzidine tetrahydrochloride (DAB; catalog no. D5637; Sigma-Aldrich) diluted in Tris-buffered saline (TBS; 50 mM; pH 7.4), to which 0.005% of H 2 O 2 was added. The reaction was stopped and the sections mounted on gelatin-coated slide and air-dried. Sections were then dehydrated in graded alcohol, cleared in toluene and coverslipped with Permount (catalog no. SP15-500; Fisher Scientific). The TH-immunostained sections taken through the midbrain were used to assess the number of DA In each mouse, the DA lesion was also assessed using an infrared imaging system (Odyssey CLx; LI-COR Biosciences, Lincoln, NE, USA) from a 50 μ m-thick section taken at the pre-commissural level of the striatum (1.34 mm from bregma). Sections were immunostained for TH and DAT, using secondary antibodies coupled to infrared fluorescent dyes. The primary antibody against TH was the same as above (1/1000, overnight at 4 o C). The monoclonal antibody against DAT (1/1000, overnight at 4 o C, catalog no. MAB369; EMD Millipore Corporation, Billerica, USA) was raised in rat. Donkey anti-rabbit 680 (1/1000, 2 h at RT, catalog no. 926-68073; LI-COR Biosciences) and goat anti-rat 800 (1/1000, 2 h at RT, catalog no. 926-32219; LI-COR Biosciences) were used as secondary antibodies. Two solid-state diode lasers (685 nm and 785 nm) were used to excite secondary antibodies coupled to infrared fluorescent dyes. Intensity values of TH and DAT immunoreactivity were taken from six 0.16 mm 2 squares randomly placed over the striatum and from one 0.16 mm 2 square placed over the Acb.
Calbindin immunohistochemistry. In order to precisely delineate the core and the shell regions of the Acb, transverse sections adjacent to those used for stereology were immunostained for calbindin (CB). Briefly, the immunoperoxidase protocol described above was used but with a monoclonal primary antibody against CB (1/500, overnight at 4 o C, catalog no. C-9848, Sigma-Aldrich) and a biotinylated horse anti-mouse secondary antibody (1/200, 2 h at RT, catalog no. BA-2000; Vector Laboratories).
μ-opioid receptor immunohistochemistry. In order to delineate the striosomes and matrix striatal compartments, 2 transverse sections were taken at the pre-commissural and post-commissural striatal levels and immunostained for μ -opioid receptor (MOR). These sections were used to assess differences in the density of D 1 /D 2 MSNs between the 2 striatal compartments. Briefly, sections were incubated with a primary antibody against MOR 6-OHDA-lesioned mice were used for stereological estimation of the number of D 1 , D 2 and D 1 /D 2 MSNs. In each mouse, 6 sections were selected from the pre-commissural striatum. Three adjacent sections were selected at 0.26 mm from bregma and 3 others at 1.10 mm. Three more sections were taken at − 0.94 mm from bregma. These sections were used to estimate the number of D 1 , D 2 and D 1 /D 2 MSNs in the pre-commissural (0.26 mm from bregma) and post-commissural striatum (− 0.94 mm from bregma), as well as in the Acb (1.10 mm from bregma), with an unbiased stereological method using a confocal microscope equipped with a digital camera and a motorized stage controlled by a computer running the StereoInvestigator software (v. 7.00.3; MicroBrightField, Colchester, VT, USA). First, on each section, striatal contour was traced and 4 striatal sectors were delineated corresponding to a dorsolateral, dorsomedial, ventrolateral and ventromedial sectors (Fig. 4). The Acb was divided into its core and shell compartments based on CB-immunostained adjacent sections.
On each section, the striatum was entirely scanned through multiple Z-stacks using a 40x objective (NA 1.4, oil immersion, Plan-Apochromat, Zeiss) and the 488 and 568 diode lasers at 2% of power. Optical resolution of the Z-stacks was 512 × 512 pixel (pixel size = 0.10 μ m 2 ), with Z-steps of 3 μ m corresponding to the optical slice determined by the pinhole. The thickness of Z-stacks was fixed at 30 μ m and the gain setting for each channel was kept constant during the entire acquisition process.
The process leading to the estimation of the total number of D 1 and D 2 MSNs began by randomly translating a grid formed by 235 × 580 μ m rectangles on the previously acquired confocal images of the striatum and the Acb. At each intersection of the grid that fell into the sector, a counting frame measuring 157 × 157 μ m was drawn and examined. Cell bodies containing tdTomato (D 1 ) or GFP (D 2 ) that fell inside the counting frame and did not contact the exclusion lines were counted whenever they came into focus within a 12 μ m-thick optical disector placed at 9 μ m from the top of the section. An average number of 2939 ± 186 D 1 and 2477 ± 160 D 2 neurons were counted in the striatum and 1187 ± 98 D 1 and 642 ± 64 D 2 cell bodies in the Acb of each mouse, yielding coefficient of error (Gundersen, m = 1) ranging between 0.05 and 0.18.
Because less numerous, the number of striatal cells containing both tdTomato (D 1 ) and GFP (D 2 ), here called the D 1 /D 2 MSNs, was estimated with stereological parameters different from those employed to estimate the number of D 1 and D 2 MSNs. Selected transverse sections of the striatum were entirely examined with an optical disector of the same size of the one described above. An average of 599 ± 106 D 1 /D 2 cells were counted in the striatum compared to 993 ± 191 D 1 /D 2 neurons in the Acb of each mouse, yielding coefficient of error (Gundersen, m = 1) ranging between 0.08 and 0.27. The density of D 1 , D 2 and D 1 /D 2 MSNs was obtained by dividing the number of striatal cells estimated with the optical disector by the volume of striatal sectors sampled, as estimated with the Cavalieri's method 71 .
In the 4 sham-lesioned animals, the 2 transverse sections taken at the pre-commissural (0.26 mm from bregma) and post-commissural (− 0.94 mm from bregma) striatal levels that were immunostained for MOR were used to estimate the number of D 1 /D 2 MSNs in the striosomes and matrix compartments. Again, all striatal D 1 / D 2 MSNs were counted from previously acquired confocal images and the striosomes and matrix compartments were delineated by using MOR immunoreactivity.

Single-cell injections of identified MSNs.
Fine morphological changes of MSN dendritic arborization induced by DA denervation were characterized by single-neuronal injections of Lucifer yellow applied to PFA-fixed brain sections from the Drd1a-tdTomato/Drd2-EGFP transgenic mice, using a method previously described 72,73 . Sharp heat-pulled glass micropipettes filled with a 4% solution of Lucifer yellow (catalog no. L453; Life Technologies, Carlsbad, CA, USA) and containing a silver electrode connected to a computer-controlled microelectrode amplifier (Multiclamp 700 A, Axon Instruments) were inserted into 250 μ m-thick brain section kept in ice-cold PBS (0.1 M), under an epifluorescence microscope (model no. E600FM, Nikon, Tokyo). After the insertion of the micropipette into the soma of visually identified MSNs located in the dorsal striatum, a negative direct current of 5 nA was administered for 20 minutes during which MSNs were filled with the negatively charged Lucifer yellow tracer. A 488 nm filter was used to visualize GFP contained in D 2 MSNs whereas a 568 nm filter was used to detect the presence of tdTomato into D 1 MSNs. Each neurons being injected was carefully inspected with both filters in order to determine content in GFP and/or tdTomato with a 40X water immersion objective (NA 0.80). All injected MSNs were located in the dorsal striatal region, at the commissural level, and restricted to 0.02-0.26 mm in the anteroposterior axis relative to the bregma, 1-3 mm in the mediolateral axis and 2-3 mm in the dorsoventral axis, according to the stereotaxic mouse brain atlas 70 .
The 250 μ m-tick sections containing injected striatal neurons were mounted on glass slides and coverslipped with a fluorescence mounting medium (S3023; Dako, Mississauga, Ontario, Canada). Z-stack of Lucifer yellow-filled neurons were obtained from the confocal microscope using a 405 nm diode laser and a 63x oil immersion objective (NA 1.4, Plan-Apochromat, Zeiss). The pixel size was 0.001 μ m 2 whereas the optical slicing was 0.3 μ m. A tiling process was used when dendritic arborization extend beyond the field of view.

Morphological analysis of injected MSNs.
Analyses of dendritic arborization and spine density were performed using the freely available NeuronStudio software (CNIC, Mount Sinai School of Medecine, New York, NY, USA). The entire somatodendritic domains of injected neurons were carefully reconstructed using maximum intensity projection in NeuronStudio software, as previously described 74 . By using a combination of automatic spine detection from NeuronStudio software followed by a careful examination of individual spine, we were able to provide faithful three-dimensional reconstructions of somatodendritic domains of injected MSNs. From these reconstructions, Sholl analyses were performed on individual MSNs of the D 1 , D 2 and D 1 /D 2 types. Measurements of the size of the cell body were conducted by using the ImageJ software (Version 1.48). In this software, maximum intensity projections of Z-stacks were generated and diameters of cell bodies were measured. Dendritic arborization analyses and size measurements were performed on 19 D 1 , 27 D 2 , 20 D 1 /D 2 Lucifer-yellow-injected MSNs from 4 sham-lesioned animals and on 18 D 1 , 17 D 2 and 22 D 1 /D 2 MSNs from 19 6-OHDA-lesioned mice.
Statistical analysis. Statistical differences of neuronal densities between different striatal and Acb regions were assessed using Kruskal-Wallis non-parametric statistical test. Differences between sham and 6-OHDA-lesioned mice regarding neuronal densities and spontaneous rotations were assessed with Mann-Whitney non-parametric test. Variations in immunoreactivity for TH and DAT between sham and 6-OHDA-lesioned mice were detected using the Mann-Whitney statistical test. A Sholl analysis with 15 μ m increments was used to determine the spine density and statistical differences for dendritic morphology (number of branching points, spine density and dendritic length) and for cell body diameters between the two experimental groups and the three types of MSNs were assessed with the two-way analysis of variance (ANOVA) followed by a Bonferroni multiple comparison test. All statistical tests were performed using GraphPad Prism software (v. 6.01; GraphPad Software, San Diego, CA, USA). Mean and standard error of the mean are used throughout the text as central tendency and dispersion measure, respectively.